phonon-lifetimes

Compute phonon power spectra, frequencies and lifetimes directly from molecular-dynamics (MD) atomic positions, for cubic cells — no perturbation theory required. Method and results are described in the accompanying publication: Phys. Rev. Lett. 123, 235501 (2019).

Why

Phonon lifetimes are usually derived from perturbative force-constant expansions. This script instead extracts them non-perturbatively from the velocity/position autocorrelation of a plain MD trajectory: it Fourier-transforms the atomic motion per q-point, fits the resulting power-spectrum peaks, and reports frequencies and linewidths (inverse lifetimes) — including all anharmonic effects sampled by the MD.

Requirements

• Python 3 (Python 2 also works — the script predates the split)
• numpy, scipy, pandas (and matplotlib for optional plotting)

pip install numpy scipy pandas matplotlib

Get it

git clone https://github.com/glensk/phonon-lifetimes
cd phonon-lifetimes

Usage (TiN example included)

cd example_TiN
tar -xf POSITIONs.tar.bzip2
../phonon_lifetimes.py -fftpy -atoms 216 -ps -v --peaks 2 -dt 1 -sc 3 -a 4.24 -pd -write_ps_fitted -write_full_ps -write_smooth_in

Key options (full list: ./phonon_lifetimes.py -h):

Option	Meaning
-sc	supercell repetitions
-atoms	number of atoms in the cell
-dt	MD timestep in fs
-a	lattice constant in Å
-ps	compute the power spectrum
--peaks	number of peaks to fit per q-point

The run writes the (smoothed/fitted) power spectra per q-point, e.g. example_TiN/ps_smooth/ps_space_fft_1_1_1_12042.dat:

Citation

If you use this code in your research, please cite:

A. Glensk et al., Phys. Rev. Lett. 123, 235501 (2019). doi:10.1103/PhysRevLett.123.235501

License

MIT